Outline
- What makes a good graph?
- What makes a good series of graphs?
- What graphical elements are appropriate for different types of data?
Making effective visualizations
Graph from last time
- What makes this a good graph?
- What makes this a bad graph?

Making effective visualizations
What makes a good graph?
- Highlights the most important information
- Reduces visual clutter (remove unnecessary backgrounds, lines, etc)
- Clear, legible elements (text is big enough, colors/shapes are distinguishable)
- Graphical elements map onto conceptual elements (graph mean +/- SE if showing inferential statistics, don’t show min/max if it’s not relevant)
- Graph does not mislead the audience (sensible axes, doesn’t hide individual data/variability)
- Scaffolds for interpreting figure (annotations, reference lines, scales)
How about this one?

How about this one?

How about this one?

How about this one?

How about this one?

How about this one?

How about this one?

Making effective sets of visualizations
- What makes a good series of graphs in a paper?
Making effective sets of visualizations
What makes a good series of graphs in a paper?
- Consistent scaling across graphs
- Same mappings for shape/color/line styles across graphs
- Consistent graphical styles (font sizes, line weights, capitalization, etc)
How about this one?

How about this one?

How about this one?

What graphical elements are appropriate for different types of data?

What graphical elements are appropriate for different types of data?

What graphical elements are appropriate for different types of data?
What graphical elements are appropriate for different types of data?
- Relationship
- Use other aesthetics – size, color, lines – to represent subgroups

What graphical elements are appropriate for comparisons?
- Comparison
- Bar/column chart, line and scatter, point range, boxplot
- When to use a line vs bar?
What graphical elements are appropriate for comparisons?
- Comparison
- Bar typically means between-subject data
- Line typically means w/in subject data (data over time/treatments)
- Point range is a better stand in for a bar chart:

Representing error/distribution in comparison graphs
What are the pros/cons of including individual data?
Representing error/distribution in comparison graphs
What are the pros/cons of including individual data?
- Pros?
- Most transparent
- Shows distribution, not just mean/sd
- Cons?
- Does not always highlight the comparison
- Raises questions about semi-outliers, influential points
- Not always possible if there’s a lot of data
Plotting individuals

Plotting groups
